請用此 Handle URI 來引用此文件:
http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/90058
完整後設資料紀錄
DC 欄位 | 值 | 語言 |
---|---|---|
dc.contributor.advisor | 葉彌妍 | zh_TW |
dc.contributor.advisor | Mi-Yen Yeh | en |
dc.contributor.author | 張烱郁 | zh_TW |
dc.contributor.author | Chiung-Yu Chang | en |
dc.date.accessioned | 2023-09-22T17:14:20Z | - |
dc.date.available | 2023-11-09 | - |
dc.date.copyright | 2023-09-22 | - |
dc.date.issued | 2023 | - |
dc.date.submitted | 2023-08-09 | - |
dc.identifier.citation | Yoshua Bengio, Nicholas Léonard, and Aaron Courville. Estimating or propagating gradients through stochastic neurons for conditional computation, 2013.
Ziheng Chen, Fabrizio Silvestri, Jia Wang, He Zhu, Hongshik Ahn, and Gabriele Tolomei. Relax: Reinforcement learning agent explainer for arbitrary predictive models. In Proceedings of the 31st ACM International Conference on Information Knowledge Management, CIKM ’22, page 252–261, New York, NY, USA, 2022. Association for Computing Machinery. I. Csiszar. i-divergence geometry of probability distributions and minimization problems. The Annals of Probability, 3(1):146–158, 1975. Dheeru Dua and Casey Graff. UCI machine learning repository, 2017. Daniel M. Hausman and James Woodward. Independence, invariance and the causal markov condition. British Journal for the Philosophy of Science, 50(4):521–583, 1999. Marco F. Huber, Tim Bailey, Hugh Durrant-Whyte, and Uwe D. Hanebeck. On entropy approximation for gaussian mixture random vectors. In 2008 IEEE International Conference on Multisensor Fusion and Integration for Intelligent Systems, pages 181–188, 2008. Eric Jang, Shixiang Gu, and Ben Poole. Categorical reparameterization with gumbelsoftmax, 2017. Kentaro Kanamori, Takuya Takagi, Ken Kobayashi, and Hiroki Arimura. Dace: Distribution-aware counterfactual explanation by mixed-integer linear optimization. In Christian Bessiere, editor, Proceedings of the Twenty-Ninth International Joint Conference on Artificial Intelligence, IJCAI-20, pages 2855–2862. International Joint Conferences on Artificial Intelligence Organization, 7 2020. Main track. Kentaro Kanamori, Takuya Takagi, Ken Kobayashi, Yuichi Ike, Kento Uemura, and Hiroki Arimura. Ordered counterfactual explanation by mixed-integer linear optimization. Proceedings of the AAAI Conference on Artificial Intelligence, 35(13):11564–11574, May 2021. Amir-Hossein Karimi, Gilles Barthe, Borja Balle, and Isabel Valera. Model-agnostic counterfactual explanations for consequential decisions. CoRR, abs/1905.11190, 2019. Diederik P. Kingma and Jimmy Ba. Adam: A method for stochastic optimization, 2017. S. L. Lauritzen and D. J. Spiegelhalter. Local computations with probabilities on graphical structures and their application to expert systems. Journal of the Royal Statistical Society: Series B (Methodological), 50(2):157–194, 1988. Alessandro Magrini, Stefano Di Blasi, and Federico Mattia Stefanini. A conditional linear gaussian network to assess the impact of several agronomic settings on the quality of tuscan sangiovese grapes. Biometrical Letters, 54(1):25–42, 2017. Divyat Mahajan, Chenhao Tan, and Amit Sharma. Preserving causal constraints in counterfactual explanations for machine learning classifiers. CoRR, abs/1912.03277, 2019. Warwick Masson, Pravesh Ranchod, and George Konidaris. Reinforcement learning with parameterized actions. Proceedings of the AAAI Conference on Artificial Intelligence, 30(1), Feb. 2016. Volodymyr Mnih, Koray Kavukcuoglu, David Silver, Alex Graves, Ioannis Antonoglou, Daan Wierstra, and Martin A. Riedmiller. Playing atari with deep reinforcement learning. CoRR, abs/1312.5602, 2013. Ramaravind K. Mothilal, Amit Sharma, and Chenhao Tan. Explaining machine learning classifiers through diverse counterfactual explanations. In Proceedings of the 2020 Conference on Fairness, Accountability, and Transparency, FAT* ’20, page 607–617, New York, NY, USA, 2020. Association for Computing Machinery. Daniel Nemirovsky, Nicolas Thiebaut, Ye Xu, and Abhishek Gupta. Countergan: Generating realistic counterfactuals with residual generative adversarial nets. CoRR, abs/2009.05199, 2020. Martin Pawelczyk, Klaus Broelemann, and Gjergji Kasneci. Learning modelagnostic counterfactual explanations for tabular data. In Proceedings of The Web Conference 2020, WWW ’20, page 3126–3132, New York, NY, USA, 2020. Association for Computing Machinery. Judea Pearl. Causality. Cambridge University Press, 2 edition, 2009. F. Pedregosa, G. Varoquaux, A. Gramfort, V. Michel, B. Thirion, O. Grisel, M. Blondel, P. Prettenhofer, R. Weiss, V. Dubourg, J. Vanderplas, A. Passos, D. Cournapeau, M. Brucher, M. Perrot, and E. Duchesnay. Scikit-learn: Machine learning in Python. Journal of Machine Learning Research, 12:2825–2830, 2011. Saptarshi Saha and Utpal Garain. On noise abduction for answering counterfactual queries: A practical outlook. Transactions on Machine Learning Research, 2022. Skipper Seabold and Josef Perktold. statsmodels: Econometric and statistical modeling with python. In 9th Python in Science Conference, 2010. Alexander A. Sherstov and Peter Stone. Function approximation via tile coding: Automating parameter choice. In Jean-Daniel Zucker and Lorenza Saitta, editors, Abstraction, Reformulation and Approximation, pages 194–205, Berlin, Heidelberg, 2005. Springer Berlin Heidelberg. Shohei Shimizu, Takanori Inazumi, Yasuhiro Sogawa, Aapo Hyvärinen, Yoshinobu Kawahara, Takashi Washio, Patrik O. Hoyer, and Kenneth Bollen. Directlingam: A direct method for learning a linear non-gaussian structural equation model. Journal of Machine Learning Research, 12(33):1225–1248, 2011. Sahil Verma, John P. Dickerson, and Keegan Hines. Counterfactual explanations for machine learning: A review. CoRR, abs/2010.10596, 2020. Sahil Verma, Keegan Hines, and John P. Dickerson. Amortized generation of sequential algorithmic recourses for black-box models. Proceedings of the AAAI Conference on Artificial Intelligence, 36(8):8512–8519, Jun. 2022. Sandra Wachter, Brent D. Mittelstadt, and Chris Russell. Counterfactual explanations without opening the black box: Automated decisions and the GDPR. CoRR, abs/1711.00399, 2017. Ermo Wei, Drew Wicke, and Sean Luke. Hierarchical approaches for reinforcement learning in parameterized action space. CoRR, abs/1810.09656, 2018. Ronald J. Williams. Simple statistical gradient-following algorithms for connectionist reinforcement learning. Mach. Learn., 8(3–4):229–256, may 1992. [31] Jiechao Xiong, Qing Wang, Zhuoran Yang, Peng Sun, Lei Han, Yang Zheng, Haobo Fu, Tong Zhang, Ji Liu, and Han Liu. Parametrized deep q-networks learning: Reinforcement learning with discrete-continuous hybrid action space, 2018. | - |
dc.identifier.uri | http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/90058 | - |
dc.description.abstract | 近期於「反事實解釋」(Counterfactual Explanation, CE) 研究探索了在保持變數因果關係的狀況下,改變分類器輸出的輸入的變數過程。我們進一步研究如何在必須變動內部變數以改變分類器輸出時,保留變數間的因果關係。具體而言,我們提出了一種名為 PoliCE 的基於強化學習的演算法,透過迭代生成跨越決策邊界所需要的每一步 (調整變數的動作)。PoliCE 找出每個內部變數在父變數給定時的可變動性,並將對其的變動分解為主動變動和固有因果效應。此外,它保證了對分類器的少量存取,因此在保留特徵因果關係的同時,可以非常高效。實驗結果顯示,PoliCE 在包括數值和類別變數的合成和真實數據集上,比過去的方法在多項指標中表現更好,尤其在保持變數間的因果關係及效率上提升顯著。 | zh_TW |
dc.description.abstract | Recent studies of Counterfactual Explanation (CE) explore the perturbation process of input features to change a classifier’s output in awareness of the causal relations among features. We further study how to preserve the inherent feature causality when the perturbation on endogenous features is necessary to changing the classifier output. Specifically, we propose PoliCE, a reinforcement learning-based algorithm to iteratively generate every step (action of tuning features) along the way to crossing the decision boundary. PoliCE finds out the perturbability of each endogenous feature given its parent features and decomposes the perturbation on it into active action and inherent causal effect. It guarantees a small number of accesses to the classifier, thus making it very efficient while preserving the feature causality. Extensive experiment results show that PoliCE outperforms the baselines on both synthetic and real datasets with both numerical and categorical features, especially in causality preservation and efficiency. | en |
dc.description.provenance | Submitted by admin ntu (admin@lib.ntu.edu.tw) on 2023-09-22T17:14:20Z No. of bitstreams: 0 | en |
dc.description.provenance | Made available in DSpace on 2023-09-22T17:14:20Z (GMT). No. of bitstreams: 0 | en |
dc.description.tableofcontents | 誌謝i
摘要iii Abstract iv Contents v List of Figures vii List of Tables viii 1 Introduction 1 2 Preliminaries and related work 5 2.1 Problem statement . . . . . . . . . . . . . . . . . . . . . . . . . . . 5 2.2 Related work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8 3 Methodology 10 3.1 Conditional perturbability loss (P loss) for causality in CE . . . . . . 11 3.2 The RL agent . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15 3.2.1 Notation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15 3.2.2 The MDP specification . . . . . . . . . . . . . . . . . . . . . 15 3.2.3 A stable and efficient policy network . . . . . . . . . . . . . 16 3.3 Causal Propagation (CP) process . . . . . . . . . . . . . . . . . . . . 20 3.4 Obtaining PoliCE . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21 4 Evaluation 23 4.1 Evaluation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23 4.1.1 Metrics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23 4.1.2 Datasets, baselines and experiment setup . . . . . . . . . . . 24 4.1.3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25 5 Ablation Study 28 5.1 P loss versus previous causal proximity . . . . . . . . . . . . . . . . 28 5.2 The impact of P loss and CP process . . . . . . . . . . . . . . . . . . 29 6 Conclusion 32 References 33 Appendices 38 A Experimental details 39 A.1 Metrics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39 A.2 Datasets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41 A.3 Experimental resources . . . . . . . . . . . . . . . . . . . . . . . . . 42 A.4 Algorithm-specific setting . . . . . . . . . . . . . . . . . . . . . . . 43 | - |
dc.language.iso | en | - |
dc.title | PoliCE: 利用策略網路高效生成基於通用因果模型的反事實解 | zh_TW |
dc.title | PoliCE: Policy Network for Efficient Counterfactual Explanation over General Causal Models | en |
dc.type | Thesis | - |
dc.date.schoolyear | 111-2 | - |
dc.description.degree | 碩士 | - |
dc.contributor.coadvisor | 林守德 | zh_TW |
dc.contributor.coadvisor | Shou-De Lin | en |
dc.contributor.oralexamcommittee | 林軒田;林智仁 | zh_TW |
dc.contributor.oralexamcommittee | Hsuan-Tien Lin;Chih-Jen Lin | en |
dc.subject.keyword | 反事實解釋,因果模型,強化學習,深度學習,策略網路, | zh_TW |
dc.subject.keyword | Counterfactual Explaination,Causal Model,Reinforcement Learning,Deep Learning,Policy Network, | en |
dc.relation.page | 43 | - |
dc.identifier.doi | 10.6342/NTU202303272 | - |
dc.rights.note | 同意授權(全球公開) | - |
dc.date.accepted | 2023-08-10 | - |
dc.contributor.author-college | 電機資訊學院 | - |
dc.contributor.author-dept | 資料科學學位學程 | - |
顯示於系所單位: | 資料科學學位學程 |
文件中的檔案:
檔案 | 大小 | 格式 | |
---|---|---|---|
ntu-111-2.pdf 此日期後於網路公開 2028-08-07 | 1.43 MB | Adobe PDF |
系統中的文件,除了特別指名其著作權條款之外,均受到著作權保護,並且保留所有的權利。